Writing tests

This document describes how to write tests for Kraken tasks.

Use Pytest fixtures

The kraken-core package provides the kraken_ctx: Context and kraken_project: Project fixtures that you can use in your unit tests. They provide a fresh kraken.core.Context and kraken.core.Project, respectively, each pointing to a temporary directory that is automatically cleaned up after the test. The kraken_project fixture is automatically the Context.root_project of the kraken_ctx.

Keep it simple

You can do the whole shebang of running kraken_ctx.execute([":mytask"]) from inside a unit test, but if you want to test only the behaviour of a single task, it is usually better to just call task.execute() directly. Note that if your task does rely on it's task.finalize() and task.prepare() method being called, make sure you do that as well.

Use output properties

Checking the status returned by task.execute() is a good way to see if the task returned the result you expected, but sometimes you may want to get some finer grained information about what happened during the execution. It often makes sense to export that information form the task's execute() method as output properties. In many cases, it is also convenient for users of the task to have access to these results.

class MyCustomTask(Task):
    some_input: Property[str]
    some_output: Property[list[str]] = Property.output()

    def execute(self) -> TaskStatus:
        self.some_output.set([self.some_input.get()])
        return TaskStatus.succeeded()

And then in the unit test:

def test__MyCustomTask__can_do_cool_things(kraken_project: Project) -> None:
    task = kraken_project.task("mytask", MyCustomTask)
    task.some_property.set("foobar")
    status = task.execute()
    assert status == TaskStatus.succeeded()
    assert task.some_output.get() == ["foobar"]